Home » Blog » Artificial Intelligence in Pharma: Software Implementation and Validation According to New EU Regulations

Artificial Intelligence in Pharma: Software Implementation and Validation According to New EU Regulations

28 August 2025

Table of contents

AI in Pharma: Software & Validation in New EU Regulations

1. AI Regulations in the EU 2025: The Artificial Intelligence Act (AI Act)

The Artificial Intelligence Act (AI Act), which came into force in February 2025, represents the world’s first comprehensive legal framework for AI, aiming to build trust in technology across Europe. It introduces a risk-based approach, classifying AI systems by their level of potential threat.

The AI Act came into effect on August 1, 2024, with its requirements being phased in gradually. The first provisions, including the ban on “unacceptable risk” and the “AI literacy” requirement, are effective from February 2, 2025. Obligations for providers of general-purpose AI models (GPAI) come into effect on August 2, 2025, though the finalization of the GPAI Code of Practice has been delayed until August 2025. Most provisions, including those concerning “high-risk” systems, will be implemented by August 2, 2026, with further implementation phases extending to summer 2027. This staggered timeline creates a complex and dynamic regulatory compliance landscape.

The AI Act defines four levels of risk: unacceptable, high, limited, and minimal. Unacceptable risk systems are strictly prohibited as they pose a clear threat to safety and fundamental rights (e.g., subliminal manipulation, social scoring, untargeted facial scanning). Provisions regarding penalties for violations of Article 5 come into force on August 2, 2025.

High-risk systems are those that pose a significant risk to health, safety, or fundamental rights. This includes AI used as safety components of products covered by EU harmonization legislation (e.g., in medicine) or listed in Annex III, unless they do not pose a significant risk. High-risk systems are subject to stringent obligations, such as adequate risk assessment, high data quality, activity logging, detailed documentation, clear user information, human oversight, and a high level of robustness, cybersecurity, and accuracy.

A key requirement of the AI Act is also “AI literacy.” From February 2, 2025, providers and users of AI systems must ensure their personnel possess a “sufficient level of AI literacy.” This requirement applies to all AI systems, not just high-risk ones, and includes the ability to assess legal and ethical implications and critically interpret results.

The AI Act is a horizontal framework designed to coexist with sectoral law. There is a need for clarity on the extent to which the general principles of the AI Act will regulate the use of AI by pharmaceutical companies, especially in the context of high-risk systems. The European Medicines Agency (EMA) and the Heads of Medicines Agencies (HMA) are actively working on their own guidelines for AI in the medicinal product lifecycle, indicating the need for specific industry regulations.

2. Practical Aspects of AI Implementations in Pharmaceutical Manufacturing: EudraLex Annex 22 Guidelines

Within the general framework of the AI Act, the pharmaceutical sector receives more detailed guidance through the update of EudraLex Volume 4. The revised Annex 11 concerning computerized systems and the entirely new Annex 22 dedicated to artificial intelligence are of crucial importance.

Revised Annex 11 – Computerised Systems strengthens the requirements for managing the lifecycle of computerized systems, emphasizing the comprehensive application of Quality Risk Management (QRM) principles at all stages. Controls related to ensuring data integrity, audit trails, electronic signatures, and system security have been clarified. The New Annex 22 – Artificial Intelligence establishes specific requirements for the use of AI and machine learning in the manufacture of active substances and medicinal products.

Scope of Application: Annex 22 applies to computerized systems where AI models are used in critical applications, i.e., those with a direct impact on patient safety, product quality, or data integrity, e.g., for data prediction or classification. This specifically concerns machine learning (AI/ML) models that gain functionality through training on data.
- Key Limitations and Exclusions: Annex 22 has very precise limitations. It applies exclusively to static models (non-adaptive during use) and deterministic models (identical inputs always yield identical outputs). Dynamic models (continuously learning) and probabilistic models (identical inputs may not yield identical results) should not be used in critical GMP applications. Furthermore,
- Generative AI and Large Language Models (LLMs) are explicitly excluded from critical GMP applications. If these models are used in non-critical applications, qualified and trained personnel must ensure their outputs are appropriate, implying a “human-in-the-loop” (HITL) approach.
General Principles: Close collaboration is required among all involved parties (Subject Matter Experts (SMEs), QA, data scientists, IT) during algorithm selection, training, validation, testing, and operations. Personnel must be appropriately qualified. Full documentation of all activities must be available and reviewed. All activities must be implemented based on the risk to patient safety, product quality, and data integrity.
Intended Use: The intended use of the model and its specific tasks should be described in detail, based on in-depth process knowledge. This includes characterizing input data and identifying limitations.
Acceptance Criteria: Appropriate test metrics must be defined to measure model performance (e.g., confusion matrix, sensitivity, specificity, accuracy, precision, and/or F1 score). Acceptance criteria must be at least as high as the performance of the replaced process.
Test Data: Test data must be representative of and extend the full sample space of the intended use. They should be stratified, cover all subgroups, and reflect limitations. The test dataset must be sufficiently large to calculate metrics with appropriate statistical confidence. Labeling of test data must be verified.
Independence of Test Data: Technical and/or procedural controls must ensure the independence of test data, meaning that data used for testing cannot be used during model development, training, or validation.
Execution of Tests: Tests must ensure that the model is suitable for its intended use and “generalizes well.” A prepared and approved test plan is required. Any deviations must be documented and justified.
Explainability: Systems must log features in the test data that contributed to classification or decisions. Feature attribution techniques (e.g., SHAP, LIME) or visual tools should be used.
Confidence: The system should log the model’s confidence score for each result. Low confidence scores should be flagged as “undecided.”
Operations (Continuous Use): The model, system, and process must be under change control. Regular monitoring of model performance and input sample space (data drift) is required.

3. AI in Drug Manufacturing: Applications and Benefits

The integration of artificial intelligence in the pharmaceutical industry is leading to significant transformations in drug discovery and development, as well as pharmaceutical sector management. AI streamlines every stage, from drug discovery to clinical trials, manufacturing, and supply chain management.

In drug discovery and design, AI accelerates the analysis of vast datasets, identifies molecular targets, and predicts drug-target interactions, reducing time and costs. It enables virtual screening of chemical libraries, proposes new structures (de novo drug design), and optimizes drug candidates.

AI support in clinical trials is equally significant. AI systems shorten the duration of clinical trial cycles by using predictive models to identify relevant information in real-world data (RWD). AI helps in more effective patient matching for studies and in their design. An important innovation is the use of digital twins – virtual patient models that simulate individual responses to therapies.

In production processes, AI is revolutionizing many aspects:

Process Automation: Automating processes, AI streamlines production, ensuring consistency in repetitive operations.
Predictive Maintenance: Continuous monitoring of production operations allows AI to identify the need for part replacement or repair before it halts operations.
Waste Reduction: AI assists in analyzing drug batches to determine where improvements can be made. AI-powered quality control systems can detect early defects, reducing waste by up to 25%.
Production Scheduling: AI optimizes schedules, minimizing changes, enabling just-in-time production, and maximizing delivery efficiency.
Anomaly Detection and Digital Factory Twin: Combining anomaly detection with digital twins enables the identification and replication of the “golden batch,” minimizing deviations.
Demand Forecasting and Inventory Management: AI transforms demand forecasting and inventory management, providing more accurate forecasts.
Smart Logistics and Supply Chain: AI optimizes routes, reducing costs, delivery time, and emissions, and improves information flow and collaboration.

The applications of AI extend throughout the entire pharmaceutical product lifecycle, from research and development, through production, to logistics and personalized medicine. The success of AI implementation in pharma is inextricably linked to a company’s data management maturity.

4. Ensuring Pharmaceutical Product Quality with Artificial Intelligence

Ensuring the quality and safety of pharmaceutical products is of paramount importance. Artificial Intelligence (AI) emerges as a transformative force, capable of redefining the landscape of quality control in pharmaceuticals.

One of AI’s most significant contributions in QC laboratories is its ability to handle and interpret colossal amounts of data. AI algorithms, particularly machine learning models, excel at processing complex datasets, uncovering hidden correlations, and providing actionable insights. This predictive analytics capability shifts quality control from a reactive to a proactive function, allowing laboratories to anticipate issues before they escalate. For example, AI can analyze spectroscopic data to predict critical quality attributes or forecast the probability of batch non-compliance.

AI, through computer vision and deep learning, is revolutionizing visual inspection, providing highly accurate and consistent automated inspection capabilities. AI-powered vision systems offer automated detection of subtle defects with greater speed and accuracy than human inspectors.

AI enhances data integrity by automating data collection, reducing manual entry errors, and applying algorithms to detect anomalies or inconsistencies in datasets. It can also provide continuous monitoring of data streams for compliance with GxP principles.

Additionally, AI improves visibility and control throughout the supply chain, from supplier qualification to the distribution of the finished product, mitigating risks associated with counterfeit drugs and low-quality materials. It also contributes to reducing the potential for human error, which is a major cost driver in pharmaceutical manufacturing.

5. AI Implementation and Validation According to New Guidelines: Practical Aspects

The implementation and validation of AI systems in pharmaceuticals require an integrated approach, combining the general principles of the AI Act, the reinforced requirements of Annex 11, and the specific guidelines of Annex 22. Annex 11 provides the foundation for managing the lifecycle of computerized systems, while Annex 22 adds AI-specific layers.

Quality Risk Management (QRM) principles must be comprehensively applied at all stages of the AI model’s lifecycle: from algorithm selection, through training, validation, testing, to operations.

Key stages of AI model validation, detailed in Annex 22, include:

Definition of Intended Use: A detailed description of the model and its tasks, based on in-depth knowledge of the process into which it is integrated.
Establishing Acceptance Criteria: Defining appropriate test metrics and acceptance criteria, which should be at least as high as the performance of the replaced process.
Rigorous Test Data Management: Test data must be representative, stratified, sufficiently large, and have verified labeling. The independence of test data from training/validation data is crucial.
Test Execution and Documentation: Tests must ensure that the model is suitable for its intended use and “generalizes well.” An approved test plan is required, and any deviations must be documented.
Explainability and Confidence: Systems should record features that contributed to decisions (e.g., SHAP, LIME) and log the model’s confidence score for each result. Low confidence scores should be flagged as “undecided.”
Continuous Monitoring and Change Control: The model and system must be under change control. Model performance and input data sample space must be regularly monitored to detect data drift.

In the context of human oversight (“Human-in-the-Loop” – HITL), the human role remains crucial. For systems where testing effort has been reduced, or in non-critical applications for Generative AI/LLM, consistent review and/or testing of each model output by an operator is required.

Practical challenges arise from the limitations of Annex 22. Companies must accurately classify their AI systems to ensure that only static and deterministic models are used in critical GMP applications.

The table below provides a practical checklist and guide for validation specialists, systematizing the detailed requirements of Annex 22.

Table 1: Key Validation Requirements for AI Models in Critical GMP Applications (based on Draft Annex 22)

Aspekt Walidacji	Wymóg (na podstawie Annex 22)	Kluczowe Rozważania/Przykłady	Odpowiedzialność (wg Annex 22)
1. Zamierzone Zastosowanie	Szczegółowy opis modelu i jego zadań; charakterystyka danych wejściowych, ograniczeń.	Pomoc lub automatyzacja; podział na podgrupy; rola operatora w HITL.	Process SME
2. Kryteria Akceptacji	Zdefiniowanie metryk testowych (np. confusion matrix, sensitivity, accuracy). Wydajność modelu co najmniej równa wydajności zastępowanego procesu.	Zmienne dla podgrup; znajomość wydajności procesu zastępowanego.	Process SME
3. Dane Testowe	Reprezentatywność i rozszerzenie pełnej przestrzeni próbki; stratyfikacja, wszystkie podgrupy. Wystarczający rozmiar danych dla pewności statystycznej. Weryfikacja etykietowania.	Uzasadnienie pre-processingu i wykluczeń.	N/A (ogólne wymagania)
4. Niezależność Danych Testowych	Dane testowe nie mogą być używane w rozwoju, treningu, walidacji. Kontrole techniczne/proceduralne (dostęp, audit trail).	Zabezpieczenie danych testowych; zasada “czterech oczu”.	N/A (ogólne wymagania)
5. Wykonanie Testów	Zapewnienie, że model jest odpowiedni do zamierzonego zastosowania i “dobrze generalizuje” (wykrywanie over/underfitting).	Zatwierdzony plan testów; dokumentacja odchyleń i niepowodzeń.	Process SME (zaangażowanie w plan)
6. Wyjaśnialność (Explainability)	Rejestrowanie cech w danych testowych, które przyczyniły się do decyzji/klasyfikacji.	Użycie technik (SHAP, LIME) lub narzędzi wizualnych (mapy ciepła); przegląd cech.	N/A (wymóg systemowy)
7. Pewność (Confidence)	Logowanie wyniku pewności modelu dla każdego wyniku.	Ustawienie progu; flagowanie jako “undecided” przy niskiej pewności.	N/A (wymóg systemowy)
8. Operacje (Ciągłe Użycie)	Kontrola zmian i konfiguracji. Regularne monitorowanie wydajności systemu i dryfu danych.	Ocena zmian pod kątem retestowania; procedury przeglądu ludzkiego (HITL).	N/A (wymóg operacyjny)

6. Summary Ai Software in Pharma Industry

The integration of artificial intelligence in the pharmaceutical industry is inevitable and offers enormous benefits. However, implementing these technologies requires a proactive and rigorous approach to regulatory compliance. It is crucial to understand and implement the requirements stemming from both general legal frameworks, such as the Artificial Intelligence Act, and industry-specific EudraLex guidelines, particularly the updated Annex 11 and the new Annex 22.

For computerized system validation specialists, this means adapting to new standards that emphasize comprehensive risk management, data integrity (especially test data), rigorous validation (including test data independence, explainability, and model confidence), and maintaining the crucial role of human oversight. The explicit limitations on the types of AI permissible in critical GMP applications (static and deterministic models) necessitate a cautious choice of technology.

The pharmaceutical industry must be prepared for the continuous evolution of regulations and invest in developing “AI literacy” competencies among personnel. The future of AI in pharma will be shaped by the ability to innovate within clearly defined and stringent regulatory frameworks, while ensuring the highest standards of patient safety and quality.

6. How TTMS can help you leverage AI in pharmaceuticals

At TTMS, we perfectly understand how challenging it is to combine innovative AI technologies with rigorous pharmaceutical regulations. Our experts support companies in safely and legally implementing solutions that increase efficiency and maintain patient trust.

Want to take the next step? Contact us and see how we can accelerate your path to safe and innovative pharmaceuticals.

TTMS blog – the world through the eyes of IT experts

Artificial Intelligence in Pharma: Software Implementation and Validation According to New EU Regulations

The modern pharmaceutical industry is dynamically evolving with Artificial Intelligence (AI), which offers unprecedented opportunities in drug discovery, production optimization, and quality control. Implementing these technologies in a regulated environment requires strict adherence to standards. A key element in ensuring patient safety and the quality of medicinal products is the validation of computerized systems in the AI era, in line with the draft of new EU guidelines for pharma. This article discusses the latest EU regulations, including the Artificial Intelligence Act and specific EudraLex guidelines, and presents practical aspects of AI implementation and validation in pharmaceuticals. 1. AI Regulations in the EU 2025: The Artificial Intelligence Act (AI Act) The Artificial Intelligence Act (AI Act), which came into force in February 2025, represents the world’s first comprehensive legal framework for AI, aiming to build trust in technology across Europe. It introduces a risk-based approach, classifying AI systems by their level of potential threat. The AI Act came into effect on August 1, 2024, with its requirements being phased in gradually. The first provisions, including the ban on “unacceptable risk” and the “AI literacy” requirement, are effective from February 2, 2025. Obligations for providers of general-purpose AI models (GPAI) come into effect on August 2, 2025, though the finalization of the GPAI Code of Practice has been delayed until August 2025. Most provisions, including those concerning “high-risk” systems, will be implemented by August 2, 2026, with further implementation phases extending to summer 2027. This staggered timeline creates a complex and dynamic regulatory compliance landscape. The AI Act defines four levels of risk: unacceptable, high, limited, and minimal. Unacceptable risk systems are strictly prohibited as they pose a clear threat to safety and fundamental rights (e.g., subliminal manipulation, social scoring, untargeted facial scanning). Provisions regarding penalties for violations of Article 5 come into force on August 2, 2025. High-risk systems are those that pose a significant risk to health, safety, or fundamental rights. This includes AI used as safety components of products covered by EU harmonization legislation (e.g., in medicine) or listed in Annex III, unless they do not pose a significant risk. High-risk systems are subject to stringent obligations, such as adequate risk assessment, high data quality, activity logging, detailed documentation, clear user information, human oversight, and a high level of robustness, cybersecurity, and accuracy. A key requirement of the AI Act is also “AI literacy.” From February 2, 2025, providers and users of AI systems must ensure their personnel possess a “sufficient level of AI literacy.” This requirement applies to all AI systems, not just high-risk ones, and includes the ability to assess legal and ethical implications and critically interpret results. The AI Act is a horizontal framework designed to coexist with sectoral law. There is a need for clarity on the extent to which the general principles of the AI Act will regulate the use of AI by pharmaceutical companies, especially in the context of high-risk systems. The European Medicines Agency (EMA) and the Heads of Medicines Agencies (HMA) are actively working on their own guidelines for AI in the medicinal product lifecycle, indicating the need for specific industry regulations. 2. Practical Aspects of AI Implementations in Pharmaceutical Manufacturing: EudraLex Annex 22 Guidelines Within the general framework of the AI Act, the pharmaceutical sector receives more detailed guidance through the update of EudraLex Volume 4. The revised Annex 11 concerning computerized systems and the entirely new Annex 22 dedicated to artificial intelligence are of crucial importance. Revised Annex 11 – Computerised Systems strengthens the requirements for managing the lifecycle of computerized systems, emphasizing the comprehensive application of Quality Risk Management (QRM) principles at all stages. Controls related to ensuring data integrity, audit trails, electronic signatures, and system security have been clarified. The New Annex 22 – Artificial Intelligence establishes specific requirements for the use of AI and machine learning in the manufacture of active substances and medicinal products. Scope of Application: Annex 22 applies to computerized systems where AI models are used in critical applications, i.e., those with a direct impact on patient safety, product quality, or data integrity, e.g., for data prediction or classification. This specifically concerns machine learning (AI/ML) models that gain functionality through training on data. Key Limitations and Exclusions: Annex 22 has very precise limitations. It applies exclusively to static models (non-adaptive during use) and deterministic models (identical inputs always yield identical outputs). Dynamic models (continuously learning) and probabilistic models (identical inputs may not yield identical results) should not be used in critical GMP applications. Furthermore, Generative AI and Large Language Models (LLMs) are explicitly excluded from critical GMP applications. If these models are used in non-critical applications, qualified and trained personnel must ensure their outputs are appropriate, implying a “human-in-the-loop” (HITL) approach. General Principles: Close collaboration is required among all involved parties (Subject Matter Experts (SMEs), QA, data scientists, IT) during algorithm selection, training, validation, testing, and operations. Personnel must be appropriately qualified. Full documentation of all activities must be available and reviewed. All activities must be implemented based on the risk to patient safety, product quality, and data integrity. Intended Use: The intended use of the model and its specific tasks should be described in detail, based on in-depth process knowledge. This includes characterizing input data and identifying limitations. Acceptance Criteria: Appropriate test metrics must be defined to measure model performance (e.g., confusion matrix, sensitivity, specificity, accuracy, precision, and/or F1 score). Acceptance criteria must be at least as high as the performance of the replaced process. Test Data: Test data must be representative of and extend the full sample space of the intended use. They should be stratified, cover all subgroups, and reflect limitations. The test dataset must be sufficiently large to calculate metrics with appropriate statistical confidence. Labeling of test data must be verified. Independence of Test Data: Technical and/or procedural controls must ensure the independence of test data, meaning that data used for testing cannot be used during model development, training, or validation. Execution of Tests: Tests must ensure that the model is suitable for its intended use and “generalizes well.” A prepared and approved test plan is required. Any deviations must be documented and justified. Explainability: Systems must log features in the test data that contributed to classification or decisions. Feature attribution techniques (e.g., SHAP, LIME) or visual tools should be used. Confidence: The system should log the model’s confidence score for each result. Low confidence scores should be flagged as “undecided.” Operations (Continuous Use): The model, system, and process must be under change control. Regular monitoring of model performance and input sample space (data drift) is required. 3. AI in Drug Manufacturing: Applications and Benefits The integration of artificial intelligence in the pharmaceutical industry is leading to significant transformations in drug discovery and development, as well as pharmaceutical sector management. AI streamlines every stage, from drug discovery to clinical trials, manufacturing, and supply chain management. In drug discovery and design, AI accelerates the analysis of vast datasets, identifies molecular targets, and predicts drug-target interactions, reducing time and costs. It enables virtual screening of chemical libraries, proposes new structures (de novo drug design), and optimizes drug candidates. AI support in clinical trials is equally significant. AI systems shorten the duration of clinical trial cycles by using predictive models to identify relevant information in real-world data (RWD). AI helps in more effective patient matching for studies and in their design. An important innovation is the use of digital twins – virtual patient models that simulate individual responses to therapies. In production processes, AI is revolutionizing many aspects: Process Automation: Automating processes, AI streamlines production, ensuring consistency in repetitive operations. Predictive Maintenance: Continuous monitoring of production operations allows AI to identify the need for part replacement or repair before it halts operations. Waste Reduction: AI assists in analyzing drug batches to determine where improvements can be made. AI-powered quality control systems can detect early defects, reducing waste by up to 25%. Production Scheduling: AI optimizes schedules, minimizing changes, enabling just-in-time production, and maximizing delivery efficiency. Anomaly Detection and Digital Factory Twin: Combining anomaly detection with digital twins enables the identification and replication of the “golden batch,” minimizing deviations. Demand Forecasting and Inventory Management: AI transforms demand forecasting and inventory management, providing more accurate forecasts. Smart Logistics and Supply Chain: AI optimizes routes, reducing costs, delivery time, and emissions, and improves information flow and collaboration. The applications of AI extend throughout the entire pharmaceutical product lifecycle, from research and development, through production, to logistics and personalized medicine. The success of AI implementation in pharma is inextricably linked to a company’s data management maturity. 4. Ensuring Pharmaceutical Product Quality with Artificial Intelligence Ensuring the quality and safety of pharmaceutical products is of paramount importance. Artificial Intelligence (AI) emerges as a transformative force, capable of redefining the landscape of quality control in pharmaceuticals. One of AI’s most significant contributions in QC laboratories is its ability to handle and interpret colossal amounts of data. AI algorithms, particularly machine learning models, excel at processing complex datasets, uncovering hidden correlations, and providing actionable insights. This predictive analytics capability shifts quality control from a reactive to a proactive function, allowing laboratories to anticipate issues before they escalate. For example, AI can analyze spectroscopic data to predict critical quality attributes or forecast the probability of batch non-compliance. AI, through computer vision and deep learning, is revolutionizing visual inspection, providing highly accurate and consistent automated inspection capabilities. AI-powered vision systems offer automated detection of subtle defects with greater speed and accuracy than human inspectors. AI enhances data integrity by automating data collection, reducing manual entry errors, and applying algorithms to detect anomalies or inconsistencies in datasets. It can also provide continuous monitoring of data streams for compliance with GxP principles. Additionally, AI improves visibility and control throughout the supply chain, from supplier qualification to the distribution of the finished product, mitigating risks associated with counterfeit drugs and low-quality materials. It also contributes to reducing the potential for human error, which is a major cost driver in pharmaceutical manufacturing. 5. AI Implementation and Validation According to New Guidelines: Practical Aspects The implementation and validation of AI systems in pharmaceuticals require an integrated approach, combining the general principles of the AI Act, the reinforced requirements of Annex 11, and the specific guidelines of Annex 22. Annex 11 provides the foundation for managing the lifecycle of computerized systems, while Annex 22 adds AI-specific layers. Quality Risk Management (QRM) principles must be comprehensively applied at all stages of the AI model’s lifecycle: from algorithm selection, through training, validation, testing, to operations. Key stages of AI model validation, detailed in Annex 22, include: Definition of Intended Use: A detailed description of the model and its tasks, based on in-depth knowledge of the process into which it is integrated. Establishing Acceptance Criteria: Defining appropriate test metrics and acceptance criteria, which should be at least as high as the performance of the replaced process. Rigorous Test Data Management: Test data must be representative, stratified, sufficiently large, and have verified labeling. The independence of test data from training/validation data is crucial. Test Execution and Documentation: Tests must ensure that the model is suitable for its intended use and “generalizes well.” An approved test plan is required, and any deviations must be documented. Explainability and Confidence: Systems should record features that contributed to decisions (e.g., SHAP, LIME) and log the model’s confidence score for each result. Low confidence scores should be flagged as “undecided.” Continuous Monitoring and Change Control: The model and system must be under change control. Model performance and input data sample space must be regularly monitored to detect data drift. In the context of human oversight (“Human-in-the-Loop” – HITL), the human role remains crucial. For systems where testing effort has been reduced, or in non-critical applications for Generative AI/LLM, consistent review and/or testing of each model output by an operator is required. Practical challenges arise from the limitations of Annex 22. Companies must accurately classify their AI systems to ensure that only static and deterministic models are used in critical GMP applications. The table below provides a practical checklist and guide for validation specialists, systematizing the detailed requirements of Annex 22. Table 1: Key Validation Requirements for AI Models in Critical GMP Applications (based on Draft Annex 22) Aspekt Walidacji Wymóg (na podstawie Annex 22) Kluczowe Rozważania/Przykłady Odpowiedzialność (wg Annex 22) 1. Zamierzone Zastosowanie Szczegółowy opis modelu i jego zadań; charakterystyka danych wejściowych, ograniczeń. Pomoc lub automatyzacja; podział na podgrupy; rola operatora w HITL. Process SME 2. Kryteria Akceptacji Zdefiniowanie metryk testowych (np. confusion matrix, sensitivity, accuracy). Wydajność modelu co najmniej równa wydajności zastępowanego procesu. Zmienne dla podgrup; znajomość wydajności procesu zastępowanego. Process SME 3. Dane Testowe Reprezentatywność i rozszerzenie pełnej przestrzeni próbki; stratyfikacja, wszystkie podgrupy. Wystarczający rozmiar danych dla pewności statystycznej. Weryfikacja etykietowania. Uzasadnienie pre-processingu i wykluczeń. N/A (ogólne wymagania) 4. Niezależność Danych Testowych Dane testowe nie mogą być używane w rozwoju, treningu, walidacji. Kontrole techniczne/proceduralne (dostęp, audit trail). Zabezpieczenie danych testowych; zasada “czterech oczu”. N/A (ogólne wymagania) 5. Wykonanie Testów Zapewnienie, że model jest odpowiedni do zamierzonego zastosowania i “dobrze generalizuje” (wykrywanie over/underfitting). Zatwierdzony plan testów; dokumentacja odchyleń i niepowodzeń. Process SME (zaangażowanie w plan) 6. Wyjaśnialność (Explainability) Rejestrowanie cech w danych testowych, które przyczyniły się do decyzji/klasyfikacji. Użycie technik (SHAP, LIME) lub narzędzi wizualnych (mapy ciepła); przegląd cech. N/A (wymóg systemowy) 7. Pewność (Confidence) Logowanie wyniku pewności modelu dla każdego wyniku. Ustawienie progu; flagowanie jako “undecided” przy niskiej pewności. N/A (wymóg systemowy) 8. Operacje (Ciągłe Użycie) Kontrola zmian i konfiguracji. Regularne monitorowanie wydajności systemu i dryfu danych. Ocena zmian pod kątem retestowania; procedury przeglądu ludzkiego (HITL). N/A (wymóg operacyjny) 6. Summary Ai Software in Pharma Industry The integration of artificial intelligence in the pharmaceutical industry is inevitable and offers enormous benefits. However, implementing these technologies requires a proactive and rigorous approach to regulatory compliance. It is crucial to understand and implement the requirements stemming from both general legal frameworks, such as the Artificial Intelligence Act, and industry-specific EudraLex guidelines, particularly the updated Annex 11 and the new Annex 22. For computerized system validation specialists, this means adapting to new standards that emphasize comprehensive risk management, data integrity (especially test data), rigorous validation (including test data independence, explainability, and model confidence), and maintaining the crucial role of human oversight. The explicit limitations on the types of AI permissible in critical GMP applications (static and deterministic models) necessitate a cautious choice of technology. The pharmaceutical industry must be prepared for the continuous evolution of regulations and invest in developing “AI literacy” competencies among personnel. The future of AI in pharma will be shaped by the ability to innovate within clearly defined and stringent regulatory frameworks, while ensuring the highest standards of patient safety and quality. 6. How TTMS can help you leverage AI in pharmaceuticals At TTMS, we perfectly understand how challenging it is to combine innovative AI technologies with rigorous pharmaceutical regulations. Our experts support companies in safely and legally implementing solutions that increase efficiency and maintain patient trust. Want to take the next step? Contact us and see how we can accelerate your path to safe and innovative pharmaceuticals.

AI in Pharma: Reshaping the Future of Research and Development

Pharmaceutical companies worldwide are witnessing unprecedented transformations as artificial intelligence reshapes every aspect of drug development and healthcare delivery. This revolution goes way beyond simple automation; it’s fundamentally changing how we discover treatments, test them, and bring them to market. 1. The Role of Artificial Intelligence in Revolutionizing Pharma AI in pharma has evolved from experimental technology to something companies absolutely need to stay competitive. The transformation touches the entire pharmaceutical value chain, letting teams rapidly analyze complex datasets while boosting research efficiency and decision-making capabilities. This integration automates those time-consuming tasks that used to eat up so much expert time, freeing people up for higher-value innovation and strategic work. When human expertise combines with AI-driven tools, the results are consistently better than either working alone. This collaborative approach is reshaping business models, operational workflows, and organizational culture across the industry. Annual AI-driven value creation for the pharmaceutical sector reaches $350–$410 billion by 2025, showing just how massive the economic impact of widespread AI implementation really is. 1.1 Brief History: The Evolution of AI in the Pharmaceutical Industry AI’s journey in pharmaceuticals spans three decades, evolving from basic computational tools to sophisticated machine learning systems. Early applications focused on data mining and pattern recognition, then gradually advanced toward specialized uses like molecular modeling and clinical trial optimization. Recent breakthroughs in generative AI and neural networks have dramatically accelerated innovation. Advanced tools now enable unprecedented achievements in protein structure prediction and drug design. This evolution reflects a broader digital transformation trend, with increasing collaboration between technology providers, pharmaceutical companies, and regulatory agencies. Leading companies have shifted from experimental pilots to core strategic AI initiatives, embedding these technologies into their research and operational processes. This transition represents a fundamental change in how the industry approaches innovation and problem-solving. 2. Key Applications of AI in Pharma Today Modern pharmaceutical companies leverage AI across multiple functions, from drug discovery through commercialization. Machine learning accounts for 38.78% of the AI in pharma market, primarily supporting target discovery, compound screening, and safety profiling. Advanced algorithms identify novel drug targets, optimize chemical compounds, and predict therapeutic efficacy and safety profiles. These capabilities revolutionize clinical trial design and execution while improving patient recruitment, retention, and real-time monitoring. In manufacturing and supply chain operations, AI enhances quality control, predictive maintenance, and inventory management. The technology supports personalized medicine by analyzing genetic, phenotypic, and environmental data to tailor individual treatments. 80% of pharmaceutical professionals report using AI to find new drugs, signifying mainstream acceptance across research and clinical work. 2.1 Drug Discovery and Development: Innovating at Speed AI dramatically accelerates drug candidate identification by analyzing vast chemical and biological datasets. Machine learning models predict drug-target interactions and optimize molecular designs, boosting the chances of success in preclinical and clinical phases. The technology facilitates drug repurposing by identifying new indications for existing compounds, making pipelines more efficient. Generative AI enables entirely new molecule and therapeutic modality design, opening pathways to address unmet medical needs. These predictive capabilities improve drug candidate selection with higher clinical success probabilities. AI-driven platforms can cut drug discovery timelines very efficienty, representing a revolutionary advancement in pharmaceutical research. 2.1.1 Virtual Screening and Drug Design AI-powered virtual screening rapidly evaluates large compound libraries, filtering out less promising candidates while prioritizing those with the highest potential. Algorithms simulate and assess molecular interactions, reducing reliance on costly laboratory experiments. These tools support both de novo drug design and lead compound optimization, streamlining development pipelines. 2.1.2 AI-driven Molecular Modeling AI models predict three-dimensional structures of proteins and biomolecules, supporting rational targeted therapy design. Advanced neural networks explore complex molecular interactions, achieving near-experimental accuracy in protein structure prediction. This breakthrough accelerates the identification of drug targets central to diseases like Alzheimer’s and cancer. 2.2 Clinical Trials Management and Optimization AI transforms clinical trial processes through improved patient recruitment, optimized trial design, and real-time monitoring capabilities. Predictive analytics identify eligible participants more efficiently, increasing trial diversity while reducing enrollment timelines. Clinical trial completion is now up to 80% faster when AI is integrated compared to traditional methods. TTMS has firsthand experience supporting pharmaceutical companies in this area. Working with a global pharmaceutical company, TTMS introduced an AI system integrated with Salesforce CRM to automatically analyze and assess key criteria from RFPs. This solution improved the bidding process for clinical trials, enabling quicker and more accurate decision-making while optimizing resource allocation. 2.2.1 Decentralized Clinical Trials AI supports trial decentralization by facilitating remote patient monitoring, virtual assessments, and digital data collection. Enhanced data integration and analysis enable broader, more diverse participation while improving result generalizability. AI-driven platforms overcome traditional logistical barriers, making trials more accessible and efficient. 2.2.2 Predictive Analytics in Patient Recruitment Machine learning models analyze electronic health records and real-world data to match eligible patients with appropriate trials. Predictive tools anticipate patient dropouts and identify retention factors, enabling proactive interventions. These capabilities improve recruitment speed and accuracy, helping trials meet enrollment targets while reducing delays. 2.3 Enhancement of Supply Chain and Manufacturing Processes AI optimizes pharmaceutical manufacturing through automated quality control, predictive equipment maintenance, and streamlined production schedules. Predictive analytics enhance demand forecasting, inventory management, and logistics while reducing waste and ensuring timely medicine delivery. AI adoption has triggered productivity increases of 50–100% in quality control activities, demonstrating significant operational improvements. 2.3.1 Intelligent Automation in Production AI-driven automation systems monitor and adjust production parameters in real time, enhancing consistency and efficiency. Virtual assistants and robotics support routine manufacturing tasks, enabling faster and more reliable output. Novartis has implemented AI-powered analytics in its manufacturing to monitor production in real time, flagging quality issues before they escalate while reducing waste and errors. 2.3.2 Real-Time Supply Chain Management AI enables end-to-end supply chain visibility, allowing proactive identification of bottlenecks and disruptions. Real-time data analysis supports dynamic inventory management, reducing shortages and excess stock. Integration with IoT devices enhances monitoring of storage conditions and product integrity throughout distribution networks. 2.4 Personalized Medicine: Tailoring Treatments to Individual Patients AI analyzes multi-omic data, patient histories, and environmental factors to identify the best individual treatment strategies. Machine learning models predict how patients will respond to different therapies, supporting precision dosing while minimizing adverse effects. This application of artificial intelligence in pharma paves the way for more effective, targeted interventions that improve patient outcomes and satisfaction. 3. Challenges and Opportunities in Implementing AI in Pharma Pharmaceutical companies face significant challenges when implementing AI solutions alongside transformative opportunities. Success requires overcoming data, ethical, regulatory, and organizational hurdles while balancing risk mitigation with bold technology investments. 3.1 Data-Driven Hurdles and How to Overcome Them High-quality, comprehensive, and interoperable data remains critical for effective AI applications, yet data silos, inconsistencies, and biases present common obstacles. Ensuring data privacy, security, and compliance with evolving regulations is essential given the sensitive nature of health information. Model transparency and explainability are necessary to build stakeholder and regulatory trust. Organizations adopt FAIR data principles and integrate robust governance frameworks to enhance AI reliability. Combining AI insights with expert human judgment helps mitigate “black box” model limitations while ensuring responsible decision-making. TTMS addresses these challenges through comprehensive data management solutions. In one implementation, TTMS developed document validation software for a leading pharmaceutical company struggling with manual document validation processes. The automated validation system improved efficiency, reduced manual errors, and ensured regulatory compliance within their electronic document management system. 3.2 Ethical and Regulatory Concerns AI in pharmaceutical applications raises important ethical questions regarding bias, transparency, and equitable access to new therapies. Regulatory agencies are developing frameworks to evaluate AI-driven solutions, but the landscape remains complex and rapidly evolving. Both the FDA and EMA emphasize explainable AI models due to the “black box” nature of many systems. This transparency is crucial for regulatory credibility, patient trust, and scientific accountability. Risk-based regulatory approaches focus on aligning best practices across medical product categories while creating adaptable standards matching AI technology’s rapid evolution. Regulatory bodies stress robust data governance, audit trails, and documentation throughout AI model lifecycles. This ensures AI-generated evidence supporting drug approvals remains trustworthy and reproducible. Meaningful human oversight for AI-driven decisions in high-risk areas, including drug approval and clinical decision support, remains mandatory across regulatory frameworks. 3.3 The Path Forward: Opportunities for Growth and Innovation Continued AI advancements offer revolutionary potential in drug discovery, accelerated clinical development, and personalized patient care. Strategic partnerships between pharmaceutical companies and technology firms foster innovation while expanding AI capabilities. The AI in drug discovery market was valued at USD 1.72 billion in 2024 and is projected to reach USD 8.53 billion by 2030, demonstrating substantial growth opportunities. Investment in interdisciplinary talent and workforce upskilling is essential to harness AI’s full potential. The evolution of AI-driven platforms and integration with other digital technologies will create new business models and therapeutic paradigms. The industry’s commitment is further evidenced by a dynamic landscape of corporate partnerships and aggressive talent acquisition.. 4. Future Trends: AI Shaping the Future of Pharmaceuticals The next decade will see AI further embedded across pharmaceutical value chain stages, driving faster, more cost-effective, and patient-centric innovation. Ongoing advancements in generative AI, quantum computing, and emerging technology integration will reshape drug discovery, development, and delivery methods. 4.1 Trends in AI-Driven Drug Discovery by 2025 and Beyond By 2025, approximately 30% of new drugs will be discovered using AI platforms, illustrating dramatic early-stage research transformation. AI-driven platforms reduce costs by up to 40% while shortening discovery timelines from five years to 12–18 months, accelerating therapeutic innovation. AI-discovered drugs in phase 1 clinical trials show higher success rates with estimated 80–90% success for AI-developed versus 40–65% for traditionally discovered drugs, especially in anticancer research. Real-world evidence and adaptive trial designs powered by AI will become standard practice, enhancing clinical research efficiency and relevance. 4.2 The Rise of Generative AI in Molecular Design Generative AI is growing fastest, with a 43.12% CAGR, primarily for novel molecule design. These models enable creation of novel molecules and proteins not found in nature, expanding innovative therapy possibilities. The global generative AI market in chemicals was valued at $317.54 million in 2024 and is projected to reach $3.72 billion by 2034. Generative AI tools support rapid hypothesis testing, molecular optimization, and de novo drug design, accelerating biomedical discovery pace. The combination of generative AI with high-throughput screening and advanced analytics transforms molecular medicine landscapes. Over 1.2 million researchers use AlphaFold and similar models globally for protein modeling and drug lead identification. 4.3 The Integration of AI with Other Emerging Technologies (Blockchain, IoT) AI convergence with blockchain enhances data security, traceability, and transparency across pharmaceutical supply chains. Integration with IoT devices enables real-time monitoring of manufacturing processes, clinical trials, and patient health while providing actionable insights. Combined use of AI, blockchain, and IoT fosters new efficiency levels, compliance standards, and operational innovation. These integrated technologies create comprehensive digital ecosystems supporting end-to-end pharmaceutical operations while maintaining security and regulatory compliance. 5. Concluding Insights: Navigating the AI Revolution in Pharma Successful pharmaceutical industry transformation through AI requires holistic strategies encompassing technology, people, processes, and governance. Organizations must foster innovation cultures, continuous learning, and cross-functional collaboration to realize AI’s full potential. Building trust, ensuring ethical practices, and maintaining regulatory compliance remain essential for sustaining progress and public confidence. The AI in pharma market is estimated at $4.35 billion in 2025, projected to reach $25.37 billion by 2030 with a rapid 42.68% CAGR, underscoring the significant growth trajectory. 5.1 Building a Sustainable AI Strategy in Pharma Robust AI strategies align technological investments with business objectives, scientific goals, and patient needs. Scalable infrastructure, effective change management, and clear governance structures prove critical for long-term success. Ongoing evaluation and adaptation of AI initiatives ensure relevance and impact as technologies and regulations evolve. Pharmaceutical companies should integrate AI for end-to-end operational efficiency, leveraging technology to enhance manufacturing, supply chain management, and predictive maintenance. Investment in explainable AI tools builds regulatory trust while easing compliance challenges. 5.2 Tips for Pharma Companies to Embrace AI Fully Start with focused, high-impact use cases to demonstrate value and build organizational momentum. Invest in data quality, interoperability, and security to establish foundations for reliable AI applications. Prioritize upskilling and recruitment of interdisciplinary talent bridging gaps between data science, medicine, and regulatory affairs. Foster partnerships with technology innovators and academic institutions to stay at the AI advancement forefront. Maintain transparent communication and stakeholder engagement to drive adoption and trust in AI-driven solutions. Develop robust data governance and sharing frameworks while standardizing AI infrastructure and workflows. 6. How TTMS Supports Pharmaceutical Companies in Implementing AI Solutions TTMS delivers tailored AI solutions for pharma, fully aligned with industry regulations such as FDA and GDPR. These solutions support every stage of the project lifecycle — from discovery workshops and planning, through AI development and validation, to training and documentation — helping pharmaceutical companies accelerate innovation, ensure compliance, and optimize operations. We support flexible engagement models: Staff Augmentation, Team Leasing, and End-to-End project delivery. This allows clients to scale projects according to their needs — whether they need a single expert or a full project team. Our certified teams integrate advanced AI capabilities into technologies like Salesforce CRM, Adobe Experience Manager, and SAP CDC/Gigya. We specialize in areas such as predictive analytics, intelligent automation, virtual assistants, and AI-driven decision support systems. For example, in a recent project with Takeda Pharma, TTMS developed an AI‑powered system within Salesforce to automate the analysis of multi‑million‑dollar RFPs. The solution automatically extracted key tender parameters, provided a preliminary compliance assessment, and flagged offer deviations. As a result, Takeda’s Polish affiliate saw significantly faster and more precise bid processing, freeing up teams to focus on strategic decisions and higher‑value tasks — you can read the full case here. TTMS has also implemented AI‑based solutions including automated document validation tools and custom healthcare‑professional portals. These real‑world applications help our clients reduce time‑to‑market, improve compliance accuracy, and deliver better patient outcomes — proving the real value of strategic AI adoption in pharma. Have a specific challenge in mind or want to learn how AI can support your team? Contact us to discuss your needs with our specialists.

6AMLD and the New EU AML Regulation: What It Means for Business

The European Union has been steadily tightening its anti-money laundering (AML) laws. The Sixth Anti-Money Laundering Directive (6AMLD) is already in force, and an ambitious new EU AML Regulation has been adopted and will apply from July 2027. Together, these measures bring significant changes for businesses across all sectors. Below, we break down the key points and practical implications of 6AMLD and the new AML Regulation for companies operating in the EU. Understanding 6AMLD: A New Level of Enforcement and Liability 6AMLD was introduced to strengthen and harmonize AML rules across EU member states. It had to be transposed into national laws by December 2020, with firms expected to comply by June 2021. Unlike prior directives, 6AMLD focuses heavily on closing legal loopholes and ensuring criminals (and complicit businesses) face tougher consequences. The main changes under 6AMLD include: Cross-Border Crime Prosecution: 6AMLD makes it easier to prosecute money laundering that spans multiple countries. It compels EU states to cooperate more effectively and allows offenses committed in different jurisdictions to be prosecuted in a single member state. Importantly, for certain serious crimes (e.g. terrorism, trafficking, organized crime), EU countries must treat them as money-laundering predicate offenses even if the conduct isn’t illegal where it occurred. This elimination of the “dual criminality” loophole means money launderers can no longer hide behind differences in national laws. Unified Predicate Offenses: The directive defines a single harmonized list of 22 predicate offenses (underlying crimes) that constitute money laundering across all EU states. New categories like environmental crime, cybercrime, and insider trading have been added to reflect modern criminal risks. Businesses must ensure their compliance programs can detect transactions linked to any of these predicate offenses, as the risk landscape has broadened. Corporate Criminal Liability: One of the most impactful changes is that legal persons (companies and partnerships) can now be held criminally liable for money laundering. If a company fails to prevent a “directing mind” (e.g. an executive or person of authority) from engaging in money laundering, the company itself can be prosecuted. Business leaders and those in positions of control may be personally accountable for lapses in supervision or controls that enable laundering. Essentially, the burden of proof is now on the company to show it took sufficient steps to prevent money laundering within the organization. This shift greatly raises the stakes for management to maintain effective AML controls. Tougher Penalties: 6AMLD mandates harsher punishments for money laundering offenses. EU Member States must impose a minimum four-year prison term for individuals convicted of money laundering (up from the previous one-year minimum). Companies convicted of involvement can face steep fines and even sanctions like temporary bans on business, exclusion from public funding, or permanent shutdowns of business operations. These tougher penalties aim to ensure AML violations are met with “effective, proportionate and dissuasive” sanctions across the EU. Criminalizing Aiding and Abetting: The directive broadens the scope of AML offenses to include aiding, abetting, inciting, and attempting to commit money laundering. So-called “enablers” – anyone who helps or tries to help launder money – can now be prosecuted as criminals, even if they didn’t personally benefit financially. For businesses, this means employees, partners, or contractors who facilitate money laundering (even indirectly) expose themselves and the company to liability. Firms must be vigilant that they are not unwittingly assisting clients or associates in illicit schemes. Practical compliance impact of 6AMLD: With 6AMLD in force, businesses have had to tighten their AML compliance programs significantly. The extension of criminal liability to companies and managers has made it imperative to identify and fix any compliance gaps quickly. Companies should update their AML/CFT policies, procedures, and training to cover the expanded list of predicate crimes and the new offense of aiding and abetting. Internal controls and oversight mechanisms (e.g. internal audits, managerial sign-offs) need strengthening to meet the higher bar of accountability. In many cases, firms are upgrading their monitoring systems and deploying specialist RegTech solutions to better detect suspicious activity. (For example, using advanced compliance software like AML Track can help companies continuously monitor transactions and beneficial owners, ensuring no red flags are missed.) Overall, 6AMLD’s message is clear: AML compliance is no longer just a legal formality, but a core corporate responsibility. Business leaders must foster a culture of compliance, as regulators now have more power to punish organizations and individuals for AML failings. The New EU AML Regulation: One Rulebook for All Members While 6AMLD is the last of the EU’s AML directives, it has already been complemented by a far-reaching EU AML Regulation adopted in 2024. In 2024, the EU adopted an AML reform package, including a new regulation that will apply from July 2027. Unlike a directive, an EU regulation is directly applicable in all member states without the need for national legislation, creating a single harmonized rulebook for AML. This regulation (officially Regulation (EU) 2024/1624), adopted in 2024 and applying from July 2027, will replace the existing 4th and 5th AML Directives and any national variations. What does the new AML Regulation introduce? In essence, it elevates and unifies AML requirements across Europe, closing gaps and ensuring consistency. Key changes that businesses should prepare for include: Stricter Customer Due Diligence (CDD): Firms will face enhanced due diligence obligations under the AML Regulation. Companies must more precisely identify and continuously monitor the beneficial owners of their customers and business partners. This means keeping up-to-date information on who ultimately owns or controls client companies, and tracking any changes. In addition, suspicious activity reporting will be under tighter timelines – regulators are imposing a deadline of five working days for obliged entities to respond to Financial Intelligence Unit (FIU) information requests. In practice, businesses need to speed up internal investigations and reporting of suspicious transactions. Even cryptocurrency transactions face greater scrutiny: the regulation explicitly extends enhanced due diligence requirements to crypto-asset service providers, meaning crypto exchanges and similar platforms must follow the same rigorous CDD and monitoring standards as banks. Caps on Cash Transactions: Large cash transactions will be curtailed across the EU to reduce money laundering via cash. The regulation sets an EU-wide upper limit of €10,000 for cash payments in business transactions. Any payment above €10k in cash will be illegal, and member states can opt to enforce even lower national limits. Furthermore, for any cash transaction of €3,000 or more, businesses will be required to verify the customer’s identity and record the details. These measures mean that sectors dealing in high-value goods (e.g. luxury car dealers, jewelers, art sellers) will need to implement strict controls on accepting cash. Companies should update their payment policies and train staff to enforce the new cash caps, where applicable, to ensure all large payments are traceable through banks or other regulated channels. Expanded Scope of Regulated Entities: The new rules bring more businesses under AML obligations. The AML Regulation broadens the definition of “obliged entities” (those legally required to implement AML controls) to include sectors and activities that were previously outside the AML net. Notably, crypto-asset service providers, crowdfunding platforms, real estate and art intermediaries, professional football clubs and agents, and high-value goods dealers (e.g. trade in precious metals or gemstones) are explicitly added to the AML regime. Even some professions like lawyers and accountants were already covered under prior directives; now the net is cast wider to capture emerging risk areas. While there is room for exemptions for very low-risk scenarios, generally more businesses than ever before must establish AML programs. If your company operates in one of these newly included industries, you will need to set up internal procedures for customer due diligence, record-keeping, and suspicious activity reporting, if you haven’t already. Even businesses that remain outside the formal list should be aware that large, unusual transactions could still trigger scrutiny under general criminal laws. Harmonized Beneficial Ownership Rules: The new framework standardizes how companies must identify and report beneficial owners (the persons who ultimately own or control an entity). Across the EU, a beneficial owner will be uniformly defined as anyone owning 25% or more of a company’s shares or voting rights, or otherwise exercising control. Previously, some countries had slightly different thresholds (e.g. “more than 25%”); the new 25% rule is clear-cut and consistent. In high-risk sectors, the European Commission can even lower the threshold to 15%, forcing identification of any owners above that lower limit. For businesses, this means compliance teams must be diligent in collecting ownership information down to these thresholds for all clients and perhaps re-papering some existing client files to meet the new criteria. Moreover, authorities are going to actively verify beneficial ownership data: under the parallel 6th AML Directive in the package, national authorities must continuously check the accuracy of information in beneficial ownership registers and interlink these registers across Europe. A centralized European access point will allow regulators to quickly retrieve ownership info across borders. The practical upshot is that it will be much harder for true owners to hide behind complex corporate structures – and companies must ensure that the ownership information they report to regulators is correct and kept up to date. New EU AML Authority (AMLA): A major institutional change is the creation of a European Anti-Money Laundering Authority (AMLA) based in Frankfurt. From 2025, AMLA will start building its capacity, and by 2026-2027 it will be fully operational. This agency will have direct supervisory powers over certain high-risk, cross-border financial institutions (up to 40 of the largest banks and fintechs in the EU) and will coordinate supervision for the broader financial and non-financial sectors. For most businesses, AMLA’s impact will be indirect but significant: it will set unified regulatory standards (by issuing guidelines, technical standards, etc.) and ensure national regulators enforce the rules consistently. If a national supervisor is too lax, AMLA can step in. The presence of a central authority means that large multinational firms might be overseen at the European level, and even smaller companies will feel the effects of more consistent, rigorous supervision standards across the board. In short, the era of “light-touch” AML oversight in any EU country is coming to an end, leveling the playing field for compliance. International businesses should welcome the consistency – but also be prepared for closer scrutiny. Practical Implications and Next Steps for Businesses A Uniform EU Compliance Framework The new AML Regulation will create a more uniform compliance environment across all member states. For businesses operating in multiple EU countries, this is a positive change – it will simplify compliance by aligning requirements and removing the need to navigate differing national laws. Companies can develop one robust AML program and apply it EU-wide, with confidence that it meets the standard everywhere. Consistency in rules (e.g. the same due diligence standards and cash limits) should enable more efficient group-wide policies and training. As the DLA Piper legal team notes, “more uniform national laws will allow for aligned processes across the EU,” benefiting internationally active companies. Greater Responsibility and Liability On the flip side, the unified regime comes with heightened accountability. Under 6AMLD and the new regulation, regulators have more tools to enforce compliance and less tolerance for failures. Senior management and boards must treat AML as a strategic priority, since they can be held personally liable for serious compliance lapses. Businesses should establish clear lines of responsibility for AML internally – for example, appointing qualified compliance officers, providing regular reports to the board, and fostering a culture where compliance concerns are raised and addressed. The era of “check-the-box” compliance is over; regulators will expect to see that firms proactively prevent money laundering, not just react after the fact. Upgraded Procedures and Training Practically, companies should review and update their AML procedures now, rather than waiting for 2027. Both 6AMLD and the AML Regulation emphasize areas that may require new internal measures. For instance, procedures for verifying beneficial owners need enhancement to meet the continuous monitoring requirement. Employee training programs should be refreshed to cover the expanded list of predicate offenses (like environmental or cybercrime-related red flags) and the new offense of aiding and abetting, so staff know how to spot and report all forms of suspicious activity. Where cash is accepted, policies must be revised to enforce the new €10k limit and ID requirements. If your business falls into a newly obliged category (e.g. a crypto service or a luxury goods trader), you may need to build an AML program from scratch – this includes drafting a risk assessment, client due diligence procedures, record-keeping systems, and reporting protocols to your national FIU. Leverage Technology for Compliance Given the broadened scope of what needs to be monitored (more transactions, more data on ownership, shorter reporting deadlines), manual compliance processes may no longer suffice. Regulatory experts are encouraging firms to upgrade their monitoring systems and consider specialist RegTech tools to handle the increased workload. Automation and data analytics can help flag suspicious transactions across multiple predicate crime categories or detect anomalies in customer behavior more effectively than traditional methods. For example, solutions like AML Track (a dedicated AML compliance platform) can assist businesses in conducting ongoing customer due diligence, screening for risk indicators, and generating required reports efficiently. By investing in technology, companies can not only ensure they meet the new requirements but also reduce the burden on their compliance teams. Stay Ahead of Enforcement It’s worth noting that regulators are not waiting until 2027 to act. The clear direction of EU law is already toward stricter AML enforcement, and national authorities are likely to intensify supervision in the interim. In fact, the EU’s AML package explicitly signals that there will be “stricter control of existing AML obligations… even before the AML Regulation fully applies,” and urges obliged entities to use the lead time to strengthen their processes. Businesses should heed this warning by conducting thorough internal audits of their AML controls now. Remediate any weaknesses – whether it’s outdated customer verification practices, backlogs in reviewing alerts, or insufficient training – as soon as possible. Regulators will view early compliance upgrades as a good-faith effort, whereas waiting until deadlines approach could invite scrutiny or penalties. A New Era of Accountability for EU Businesses In summary, the 6AMLD and the new EU AML Regulation together herald a new era of anti-money laundering compliance in Europe. The practical impact on businesses will be substantial: companies face a broader scope of regulated activities, more stringent due diligence duties, and direct liability for missteps. Yet, these changes also bring benefits in the form of clearer rules and a level playing field across the single market. Businesses that proactively adapt – by reinforcing their compliance frameworks, training their people, and employing smart technological solutions – will not only reduce their risk of penalties but also help foster a safer and more transparent financial environment. The ultimate goal of these reforms is to protect honest enterprises and the economy at large from the harms of money laundering. For business leaders, that means compliance is not just a legal checkbox, but a vital component of corporate integrity and sustainability. Embracing these AML changes today will prepare your organization for the more unified, accountable, and resilient marketplace of tomorrow. AML Track: Supporting Businesses in the New Compliance Era Adapting to 6AMLD and the new EU AML Regulation may seem overwhelming, but technology can ease the burden. AML Track is an advanced compliance platform designed to help companies meet these heightened requirements. It automates customer due diligence, monitors transactions in real time, screens clients against sanctions lists, and generates audit-ready reports. By centralizing AML processes, AML Track not only reduces the risk of human error but also ensures organizations stay aligned with evolving EU standards. For businesses navigating stricter obligations and liability, AML Track offers a reliable way to strengthen defenses and maintain regulatory confidence. How does 6AMLD differ from the previous AML directives? 6AMLD significantly raised the stakes compared to earlier directives by introducing corporate criminal liability, expanding the list of predicate offenses, and harmonizing definitions across the EU. Unlike earlier rules that left more discretion to member states, 6AMLD closed loopholes such as “dual criminality” and required harsher sanctions. This means that both individuals and companies can face tougher penalties, and compliance must be more proactive and comprehensive. What impact will the EU AML Regulation have compared to directives? Directives require transposition into national law, which often leads to variations in implementation. The new AML Regulation, however, is directly applicable in all member states, creating a uniform set of rules across the EU. For businesses, this removes the complexity of adapting compliance programs to different national frameworks. Instead, one harmonized system will apply, which simplifies some aspects but also eliminates flexibility and excuses for non-compliance. Why is beneficial ownership such a focus in the new rules? Beneficial ownership transparency is at the heart of the EU’s fight against money laundering because criminals often hide behind complex corporate structures. The AML Regulation enforces a consistent definition of beneficial owners and requires companies to identify, monitor, and update this information continuously. This not only prevents abuse of shell companies but also increases pressure on businesses to maintain accurate, up-to-date records, which regulators can now easily cross-check through linked registers. How will AMLA change the enforcement landscape in Europe? The creation of the European Anti-Money Laundering Authority (AMLA) marks a turning point in supervision. AMLA will directly oversee the riskiest cross-border financial institutions and set harmonized standards for all obliged entities. While most companies will still report to their national supervisors, AMLA ensures consistency and can intervene if national regulators are too lenient. This means enforcement will be stricter, more uniform, and more predictable across the EU, raising the overall bar for compliance. What should businesses do now to prepare for the new AML Regulation? Companies should not wait until 2027 to start adjusting. Instead, they should review their AML policies, strengthen customer due diligence procedures, and ensure their systems can handle tighter reporting deadlines. Training staff to recognize the expanded range of predicate offenses and implementing technology-driven monitoring tools are also crucial steps. By acting early, businesses reduce the risk of penalties, build trust with regulators, and position themselves as leaders in compliance rather than followers scrambling to catch up.

AML in the Art Market: Automation for Safe and Transparent Transactions

AML in the Art Market: Automation for Safe and Transparent Transactions Did you know that criminals launder an estimated $3 billion through art each year? The global art market – worth over $65 billion annually – has long been a target for illicit finance. Fraudsters and kleptocrats have taken advantage of the art world’s secrecy and soaring prices to turn “dirty” money into legitimate assets. This article explores why art is so attractive for money laundering and how European anti-money laundering (AML) laws – particularly the 5th AML Directive – impose new duties on art galleries and auction houses. We’ll also discuss the serious risks of non-compliance and how automation can help art businesses meet their AML obligations, ensuring safer, more transparent transactions. Why the Art Market Attracts Money Launderers The art market’s allure for money launderers comes down to its unique combination of high value and opacity. A painting or sculpture can be worth millions yet is easily portable and concealable. In fact, art is often described as an “ideal playground for money laundering”. Here are some key reasons why criminals turn to art: Anonymity and secrecy: Art sales have traditionally been private, with buyers and sellers able to remain anonymous through shell companies or agents. Until recently, dealers and auction houses had no legal obligation to identify clients or report suspicious activities, allowing illicit actors to operate in the shadows. High-value, portable assets: Artworks pack immense value into a small package – a single painting can be worth tens of millions. These assets can be moved across borders or kept in offshore storage with little detection. For example, small collectibles like rare coins or antiquities can be smuggled easily, making art a convenient vehicle to transfer wealth secretly. Subjective pricing: There’s no fixed “market price” for a masterpiece – value is in the eye of the beholder. This subjectivity lets criminals manipulate prices to launder funds. They might overpay for a piece using dirty money and later sell it for a “clean” profit, or trade art at an inflated or deflated price between colluding parties to obscure the money trail. Free ports and storage: Valuable art is often stored in tax-free port warehouses (in jurisdictions like Geneva or Monaco) that offer high security and anonymity. Art can sit in a free port for years “in transit,” changing ownership on paper without ever leaving the warehouse. This makes it easy to conduct secret transactions beyond the reach of most regulators. Cash purchases and intermediaries: Traditionally, art deals could be done in cash, and auction houses often dealt with intermediaries rather than the ultimate buyer. This meant the true source of funds could be obscured. Many major auction houses historically did not ask for the identity of the actual client or ultimate beneficial owner (UBO) behind a purchase. Such gaps have been exploited by money launderers to inject illicit cash into the art trade without scrutiny. These factors have led to notorious cases where art was used to launder money. In one U.S. case, drug traffickers accepted 33 paintings as payment for narcotics and planned to resell them to “clean” the money, a scheme that landed the perpetrators in prison. As law enforcement notes, the volume of questionable transactions in the art market is noticeably higher than in other sectors. Recognizing this vulnerability, authorities worldwide have begun closing the loopholes that made art a safe haven for illicit funds. EU AML Legislation: 5AMLD and the Art Sector In the European Union, regulators responded to the art market’s money-laundering risks by extending AML laws to art businesses. The Fifth Anti-Money Laundering Directive (5AMLD), which took effect January 2020, explicitly brought art dealers, galleries, and auction houses into the scope of AML regulation. Under 5AMLD’s definition, any “persons trading or acting as intermediaries in the trade of works of art, including when carried out by art galleries and auction houses,” as well as those storing or trading art in free ports, are considered “obliged entities” when transactions exceed €10,000. In practice, this means if you operate in the EU art market and engage in high-value sales (even as a series of linked transactions), you must follow the same AML requirements as banks and other financial institutions. Crucially, EU legislation requires a risk-based approach to AML in the art sector. Galleries and auction houses must assess the risk of money laundering in each transaction and client relationship, focusing more effort on higher-risk cases (such as unusual payments or politically exposed buyers). The 4th AML Directive had already covered businesses receiving large cash payments (≥ €10,000), but 5AMLD went further by targeting the art trade’s particular vulnerabilities. In short, anonymity is no longer an acceptable norm – European law now requires art market participants to disclose buyer and seller identities and scrutinize the source of funds. It’s also worth noting the Sixth Anti-Money Laundering Directive (6AMLD), which EU member states implemented starting 2021, strengthens penalties and enforcement. 6AMLD harmonizes the definition of money laundering across the EU and imposes tougher punishments on individuals and companies involved. For example, it sets a minimum prison term (often around four years) for serious money laundering offenses and can hold companies liable for facilitating money laundering. Together, 5AMLD and 6AMLD send a clear signal: art businesses in Europe must take AML compliance seriously, or face severe consequences. Key AML Obligations for Galleries and Auction Houses Under these EU directives (and equivalent UK regulations for British art market participants), galleries and auction houses now have concrete AML duties. In practice, art businesses must establish internal compliance programs similar to those in finance. The key obligations include: Client due diligence (CDD): Verify the identity of clients and collect relevant information before completing a sale. This “Know Your Customer” process means obtaining official ID documents, proof of address, and understanding the nature of the client’s business and funds. If the client is a company or buying through an agent, the gallery must identify the ultimate beneficial owner (UBO) – the real person behind the transaction – and verify their identity. Risk assessment and ongoing monitoring: Evaluate each client and transaction for risk factors (e.g. unusually high-value purchases, payments from high-risk countries, politically exposed persons) and apply proportional scrutiny. After onboarding, continue to monitor transactions for any red flags. Large or complex transactions that lack obvious economic rationale should prompt further inquiry into the source of funds. Galleries should also pay attention to any changes in a client’s profile or behavior over time. Screening against sanctions and PEP lists: Check clients’ names against international sanctions lists and databases of politically exposed persons (PEPs) as part of due diligence. If a collector is a sanctioned individual or a high-profile political figure, enhanced due diligence and potentially rejecting the transaction may be required. Similarly, scanning for adverse media (negative news) about clients can reveal involvement in fraud, corruption, or other crimes that pose money-laundering risk. Record-keeping: Maintain detailed records of transactions, customer identification data, and the steps taken to comply with AML requirements. EU rules typically require keeping these records for at least five years. This includes copies of IDs, invoices, contracts, provenance documentation, and internal notes on risk assessments. Good record-keeping ensures transparency and is invaluable if investigators ever scrutinize a transaction. Reporting obligations: If a transaction or client activity looks suspicious or involves funds known or suspected to be criminal in origin, the business must file a Suspicious Activity Report (SAR) with the national Financial Intelligence Unit. This legal duty is akin to the reporting that banks do. Additionally, any cash transactions above certain thresholds (e.g. €10,000) should be reported when required. Prompt reporting shields the gallery/auction house from liability and aids law enforcement in tracking illicit networks. To fulfill these obligations, art market participants should also appoint an AML compliance officer and train their staff on compliance procedures. Employees must be trained to spot red flags – such as a buyer refusing to provide information, insisting on paying in cash, or using an overly complex ownership structure for a purchase. Ultimately, a culture of compliance and ethical conduct is now an expected part of the art business. By conducting proper due diligence and documentation, galleries and auctioneers not only follow the law but also help protect the integrity of the art market. Operational and Reputational Risks of Non-Compliance Failing to comply with AML laws can be disastrous for an art business. The immediate risk is legal: authorities can impose hefty fines, revoke licenses, and even pursue criminal charges if a gallery or auction house is found complicit (even unwittingly) in money laundering. Under EU rules, those involved in laundering schemes can face prison sentences – 6AMLD mandates stricter penalties, including possible minimum prison terms for serious offenses. In some jurisdictions, individuals have been sentenced to years in jail for art-related money laundering conspiracies. Regulators are actively enforcing the new rules; for example, in the UK more than 30 art businesses were fined within two years of the 2020 law for failing to register or comply with AML requirements. The message is clear: non-compliance is not treated lightly. Beyond fines and legal sanctions, consider the reputational damage that comes with an AML scandal. The art world operates on trust and reputation. If a gallery becomes known as a hub for shady transactions, it risks losing the confidence of legitimate clients, banks, and partners. Reputation loss can lead to a swift downturn in business – collectors will shy away, fearful of being tainted by association. As experts note, rebuilding trust after such damage can take years. Moreover, employees may quit and talent may be harder to attract if the company’s name is sullied. In short, the cost of non-compliance far outweighs the investment needed to build a solid AML program. By contrast, those who comply demonstrate integrity and due care, which can become a competitive advantage in an increasingly transparency-conscious market. Automation: Supporting Safe and Transparent Transactions Keeping up with AML compliance can seem daunting, especially for smaller galleries and auction houses with limited staff. This is where automation and technology-driven solutions make a difference. By digitizing and streamlining compliance workflows, art businesses can meet regulatory requirements efficiently and accurately. In fact, regulators and industry groups encourage the use of technology to strengthen AML controls in the art trade. Here’s how automation supports compliance: Digital client onboarding: Instead of relying on paper forms and manual ID checks, galleries can use secure online platforms to onboard clients. Clients can submit identification documents electronically, which can be verified instantly using AI-powered tools or databases. This not only speeds up the process but also catches fake IDs or inconsistencies more reliably. A digital audit trail is created for each customer, showing exactly when and how their identity was verified – useful evidence of compliance. Automated screening and due diligence: Compliance software can automatically screen new clients against sanctions lists, PEP lists, and watchlists in real time. It can also pull in adverse media results on a client with a click. By automating these database checks, art businesses ensure no client is overlooked and that risk-relevant information (like a buyer’s political exposure or a negative news article) is surfaced immediately. Sophisticated platforms even assign a risk score to clients based on factors like country of origin, transaction size, and profile, guiding the business on when to apply enhanced due diligence. Transaction monitoring systems: For auction houses managing numerous sales, software can monitor transactions and flag patterns that might indicate money laundering. For example, splitting a large payment into smaller ones, or rapid resales of a high-value piece, would trigger alerts. Rules can be set to catch anomalies (e.g. a purchase far above the estimated value, or a client buying art inconsistent with their known wealth profile). Automated alerts allow compliance officers to investigate timely. This kind of continuous monitoring is difficult to achieve manually, but machines excel at scanning data for irregularities 24/7. Secure record-keeping: An AML software solution provides centralized record-keeping where all client due diligence files, risk assessments, and transaction records are stored securely. Instead of shuffling through filing cabinets, compliance staff can retrieve any record in seconds. Built-in retention schedules ensure you keep data as long as legally required. In the event of an audit or inspection, having well-organized digital records dramatically reduces the effort to demonstrate compliance. It also helps in maintaining consistency – for instance, ensuring every high-value sale has an ID on file and a recorded source of funds check. By leveraging automation in these ways, galleries and auctioneers can turn a compliance burden into a business strength. Technology reduces human error and frees up staff time, allowing compliance officers to focus on analyzing truly suspicious cases rather than getting bogged down in paperwork. It also gives owners peace of mind that nothing will “slip through the cracks” – the system will flag missing information or unusual behavior automatically. In an industry where regulations are evolving, an automated solution can be updated to keep pace with new rules, ensuring continuous compliance without constant retraining. AMLTrack – AML automation tailored for the art market AMLTrack is a compliance platform developed by TTMS in partnership with the law firm Sawaryn & Partners, designed to automate key anti-money laundering processes for obliged entities, including galleries and auction houses. The system streamlines client due diligence by verifying identities, checking customers against international sanctions and PEP lists, and retrieving data from official registers (such as KRS, CEIDG, and CRBR in Poland). It also supports risk assessment, generates compliance reports, and securely archives all actions to ensure full audit readiness. By minimizing human error and reducing the burden of manual checks, AMLTrack enables art market participants to meet EU AML requirements more efficiently, safeguard their reputation, and protect their businesses from regulatory penalties. Ultimately, embracing digital AML tools helps art businesses fulfill the dual mandate of safety and transparency. It reassures clients that your gallery or auction house is a reputable, law-abiding place to do business, while making it far harder for criminals to exploit your platform. As the EU’s AML directives have shown, the era of art market opacity is ending. Galleries and auction houses that invest in compliance – and smart automation – are not only avoiding penalties, they are protecting their reputation and contributing to a cleaner, more transparent art market for all. Do all galleries and auction houses in the EU need to comply with AML regulations? Yes. Under the EU’s 5th Anti-Money Laundering Directive (5AMLD), all galleries, auction houses, and art dealers involved in transactions exceeding €10,000 must implement comprehensive AML procedures. This requirement applies to individual sales as well as multiple linked transactions totaling that amount or more. What specific AML obligations do art market businesses have under EU law? Art businesses must carry out client due diligence (CDD), verify the identity of buyers and beneficial owners, screen clients against sanctions and politically exposed persons (PEP) lists, monitor transactions for suspicious activity, maintain detailed records, and report suspicious transactions to financial authorities. What makes the art market particularly attractive for money laundering? The art market offers a combination of high-value assets, portability, privacy, and subjective valuation—ideal conditions for concealing and transferring illicit funds. Historically limited regulatory oversight and opaque transactions have further attracted criminals looking to legitimize illegal wealth. Can AML automation really help smaller galleries comply with EU regulations? Yes. Automation significantly simplifies compliance processes for galleries and auction houses of all sizes. Digital tools streamline client onboarding, automate identity checks, continuously monitor transactions, and ensure robust record-keeping, helping even small businesses meet complex regulatory requirements without needing extensive compliance teams. What happens if an art gallery or auction house does not comply with AML regulations? Non-compliance with AML rules can result in severe financial penalties, legal sanctions, and potentially criminal charges under EU laws like 6AMLD. Beyond legal consequences, businesses risk serious reputational damage, loss of client trust, reduced market opportunities, and difficulties restoring their standing within the art community.

LLM-Powered Search vs Traditional Search: 2025-2030 Forecast

When Will AI Search Overtake Google? Large Language Model (LLM)-powered assistants (like ChatGPT, Bard, and Bing Chat) are rapidly changing how people find information. This report projects when such AI-driven search will overtake traditional search engines (e.g. Google) in global consumer usage. We examine current adoption trends, growth rates, user behavior shifts, and industry forecasts to identify a “tipping point” where LLM-based search surpasses classic search in daily usage share and query volume. The focus is on 2025 through 2030, with data-driven milestones and a forecasted intersection of adoption curves around the end of the decade. Google Still Crushes AI Tools in Search Volume Traditional search engines still dominate overall query volume as of mid-2025. Google alone processes on the order of 15+ billion searches per day (well over 5 trillion annually) and maintains roughly 90% of the global search market share. By contrast, ChatGPT – the leading LLM-based assistant – handles an estimated tens of millions of “search-like” queries per day in 2025. In other words, Google Search’s daily query volume remains vastly higher – SparkToro estimated that in 2024 Google handled roughly 373× more queries than ChatGPT, and all AI-powered search tools combined made up less than 2% of the market. Even Bing (the #2 traditional engine) sees hundreds of millions of searches each day, an order of magnitude above ChatGPT’s query count. LLM-based search accounts for about 5.6% of desktop search traffic in the U.S. as of June 2025 (up from roughly half that a year earlier), according to the Wall Street Journal — still a small fraction of traditional search volume, but growing rapidly. However, the landscape is starting to shift. Google’s search traffic has continued to increase into 2025 (over 20% year-over-year growth in 2024) in part due to new AI-powered features in Search. At the same time, ChatGPT’s adoption has been explosive – it reached 100 million users within 2 months of launch (the fastest-growing consumer app ever) – and by late 2024 it was reportedly logging around 1 billion interactions per day. By early 2024, ChatGPT’s web traffic even surpassed Bing’s in volume, making it arguably the second-most used search tool on the web in some analyses. In short, Google’s lead remains enormous in absolute terms, but AI assistants are rapidly narrowing the gap from a zero baseline. Users are increasingly turning to LLM-based tools for information queries, signaling a gradual shift in the search landscape as we head further into 2025. Rapid Adoption of LLM Search Consumer uptake of LLM-based tools has been remarkably fast. A March 2025 survey found 52% of U.S. adults have now used an AI LLM (e.g. ChatGPT), signaling mainstream awareness. Among LLM users, two-thirds report using them “like search engines” for information retrieval. In other words, a significant share of the population is already turning to chatbots for search-like queries. This adoption cuts across demographics – while younger, educated users lead slightly, even 53% of U.S. adults earning under $50k have used LLMs. LLMs appear to be one of the fastest-adopted technologies in history. Several factors drive this growth: conversational convenience, always-on assistance, and rapid improvements in capability. Unlike traditional search, an LLM agent can engage in multi-turn dialogue, provide direct answers with context, and even perform tasks (coding, writing) beyond static information lookup. This versatility has led to surging usage rates. OpenAI’s ChatGPT went from launch in late 2022 to 800 million weekly active users by April 2025 – an 8× increase in just 18 months. By mid-2025 it was handling 1 billion searches per week (roughly 143 million per day) as users increasingly treat it as an information source. Other LLM-powered assistants (Anthropic’s Claude, Google’s Bard/Gemini, etc.) are also growing, though they remain much smaller than ChatGPT so far. Voice assistants are another vector accelerating AI search adoption. Globally, voice-enabled AI assistants (Siri, Alexa, Google Assistant, etc.) have proliferated – 8.4 billion voice assistants are in use by 2025, almost doubling from 4.2B in 2020. About 20-30% of consumers use voice search regularly, often for quick queries. As these voice interfaces integrate advanced LLMs, they effectively become conversational search engines, further shifting queries away from traditional typed search. The convenience of asking a question out loud and getting a spoken answer (e.g. via smartphones or smart speakers) has normalized AI-assisted search in daily life. From Search Bar to AI Chat Crucially, consumers are learning when to use LLM assistants versus a traditional engine. Studies show that 98% of ChatGPT users still also use Google – they are not abandoning one for the other outright, but rather allocating different query types to each. Simple factual or navigational queries (“weather tomorrow”, “Facebook login”) still default to Google’s quick answers. Google’s familiarity and speed make it the go-to for one-off facts or transactional searches. However, for complex, open-ended tasks – e.g. planning travel itineraries, researching a topic in depth, troubleshooting code, brainstorming – users increasingly prefer AI assistants. ChatGPT can synthesize information from multiple sources and provide a personalized, conversational response that would otherwise require many Google queries and clicks. This emerging division of labor in search is evident: users report turning to Google for quick answers but using ChatGPT for detailed explanations, creative ideas, and multi-step research. Younger demographics especially are embracing “AI-first” search habits. Nearly 80% of Gen Z have used generative AI tools, with almost half using them weekly. A majority of these young users say AI makes finding information easier (72%) and helps them learn faster. They are comfortable asking chatbots for homework help, product recommendations, or advice – queries that older users might still direct to Google or specific websites. Additionally, specialized search alternatives like TikTok (for how-tos, trends) and Reddit (for human reviews) are diverting searches from Google. In fact, “reddit” is now one of the most searched terms on Google itself, reflecting how people seek community-sourced input to validate AI or search results. All these trends indicate a broad fragmentation of search behavior: consumers are no longer relying on a single platform, but rather using a mix of AI assistants, social platforms, and traditional engines based on the context of their query. How Google Is Fighting Back with AI Facing this shift, incumbent search providers are aggressively integrating LLM technology into their products. Google launched its Search Generative Experience (SGE) in 2023-2024, augmenting search results with AI summary “Overviews”. Early results showed increased user engagement – Google’s CEO noted higher search usage and satisfaction among those using AI Overviews. Internally, Google acknowledges the landscape change: in late 2024, CEO Sundar Pichai called 2025 “critical” to address the ChatGPT threat. Google is reportedly investing $75 billion in AI to bolster its search AI capabilities, including developing its own advanced models (e.g. Gemini). The head of Google Search, Elizabeth Reid, even suggested the classic Google search bar will become “less prominent over time” as AI interfaces take center stage. Microsoft has taken a different tack – rather than defending an existing monopoly, it partnered with OpenAI to leapfrog Google. Microsoft’s $13 billion+ investment in OpenAI brought GPT-4 into Bing in early 2023, spurring a surge of interest. Within a month of adding the AI chat feature, Bing exceeded 100 million daily active users for the first time (still a single-digit share of the market, but a notable bump). Microsoft reports that roughly 1/3 of Bing’s daily users engage with AI chat and that AI features increased overall time spent on Bing. Additionally, new AI-centric search startups (Perplexity, Neeva before its pivot, etc.) have drawn significant venture funding, and OpenAI itself is exploring a dedicated AI search engine as of 2024. In China, Baidu introduced its Ernie AI chatbot into search, and other regional engines are following suit. Across the board, massive investment is flowing into AI-driven search, signaling industry consensus that LLMs are the future interface for information retrieval. AI Could Overtake Google Search by 2028 When will LLM-based search overtake traditional search? Based on current trajectories, multiple analyses converge on the late 2020s as the critical inflection period. Key data points and projections include: 2025: LLM usage still <5% of global search queries. Google remains ~90% of the market, but AI chat queries are growing exponentially. ChatGPT’s query volume is on track to reach hundreds of millions of searches per day (it hit ~143M/day by mid-2025). By 2025, over half of consumers have tried LLM search and 34% use an LLM daily or near-daily. Milestone: OpenAI’s ChatGPT crosses 1 billion weekly searches and 800M users. 2026: Inflection point begins. Gartner predicts that by 2026, traditional search engine volume will drop 25% as users turn to generative AI assistants — a shift that could mean Google’s query count peaks and starts to decline to around 10–11 billion per day (down from roughly 14 billion), while AI-powered queries continue their exponential rise. In practical terms, this could mean Google’s own query count peaking and starting to decline (~10-11B/day, down from 14B) while LLM queries continue to rise. Milestone: AI chat integrated into most search platforms (e.g. Apple potentially launches an AI search tool), and a quarter of all search queries could be handled by LLMs (per Gartner’s scenario). 2027: Early signs of parity in specific domains. Research suggests that by late 2027, AI-driven search traffic could deliver equal — or even greater — economic value to traditional search traffic, even if raw volume is lower, thanks to significantly higher conversion rates. An Ahrefs study found AI search visitors convert up to 23× better than regular search visitors, while Semrush data indicates that AI-driven traffic achieves, on average, a 4.4× higher conversion rate than traditional organic search. If these patterns hold, AI-powered channels could match Google’s business impact as early as Q4 2027. Some niche sectors may already see AI tools surpass Google in share of queries (e.g. coding help, certain research domains). In fact, early market data suggests that in areas like programming assistance, academic research, and complex product recommendations, AI-first search platforms are already capturing a majority share of queries — in some cases exceeding 60% — well before the projected 2028 tipping point. Milestone: Internal data shows AI searches overtaking traditional search for digital marketing queries by early 2028 if trends continue. 2028: Tipping point approaches. Gartner projects that by 2028, organic search traffic to websites will be down 50% or more as consumers fully embrace generative AI search. In other words, roughly half of search activity may be happening through AI assistants instead of classic search engines by 2028. Research from Semrush even predicts that AI-powered search could overtake traditional search traffic entirely by the first half of 2028 – potentially marking the crossover earlier than many industry forecasts suggest. Similarly, other market analyses suggest LLM-based platforms will capture between 30% and 50% of the search market by 2028, depending on the metric and region — with some high-engagement categories, like in-depth research or technical problem-solving, already leaning toward AI-first search dominance. Milestones: Google’s AI-driven “SGE” likely becomes the default search mode, and AI-first search engines handle an estimated 30-40% of informational queries across industries. This year is a plausible “crossover” in certain metrics (e.g. time spent or number of informational queries on AI platforms vs Google). 2030: LLM search overtakes traditional search in general consumer usage. By 2030, extrapolating current growth, AI-powered assistants are expected to handle a majority of search queries worldwide. Industry analyst Kevin Indig’s modeling (using Similarweb traffic trends) predicts ChatGPT’s traffic will surpass Google’s by around October 2030. Based on mid-2025 Similarweb data, Google Search is generating roughly 136 billion monthly visits compared to about 4 billion for ChatGPT — meaning that, to meet this forecast, AI-powered platforms would need to sustain their current double-digit monthly growth rates while Google’s traffic trends downward. In this scenario, LLM-based systems collectively would command over 50% of global search query volume by 2030, marking the definitive tipping point where AI search dominates. Google will still generate enormous query volume, but much of it may come from users asking Google’s own AI (Bard/SGE) for answers, blurring the line between “traditional” and “AI” search. Milestone: By 2030, LLM assistants become the first preference for finding information for most users – effectively “Google” becomes just one of many AI-powered or hybrid search options, rather than the default starting point. All forecasts carry uncertainty, but the consensus is that late this decade (2028-2030) will witness the crossover. By that time, LLM-based search will likely have 30-50%+ usage share, exceeding the old query-and-click model. Some optimistic scenarios even envision Google’s share dropping to ~20% by 2027 in certain verticals, with ChatGPT and others absorbing the rest. More conservative outlooks (e.g. Gartner) still see at least half of search queries shifting to AI by 2028. Our forecast aligns with these, pegging 2029-2030 as the period when AI-driven search usage definitively surpasses traditional search worldwide. What Will Speed Up (or Slow Down) AI Search Takeover? Several drivers will determine how quickly LLM search overtakes traditional search: Quality and Trust: LLMs need to continually improve accuracy and cite reliable sources. Increased trust (already ~70% of consumers trust AI results to some extent) will encourage more users to switch fully to AI for answers. Google’s integration of citations and real-time data into its AI results, as well as OpenAI’s move to connect ChatGPT to the live web, are addressing this. If by ~2025-2026 LLMs can reliably answer most factual queries with sources, users will have less need to “double-check” on Google. User Experience & Convenience: LLM assistants offer a conversational, one-stop experience (no multiple clicks), which users find appealing for complex queries. As interfaces improve (e.g. voice integration, multimodal capabilities, memory of past queries), they will attract more search share. Voice search growth also plays a role – speaking a query to an AI assistant that talks back is a natural evolution. By 2030, we expect voice and chat-based search to converge, providing instant answers on-the-go, which traditional web search can’t match for convenience. Integration into Daily Tools: AI search will become embedded in productivity apps, browsers, and operating systems. For example, Microsoft is embedding ChatGPT (via Copilot) across Office and Windows, so users can ask questions without opening a browser at all. If asking your desktop or AR glasses an question yields an immediate AI answer, the need to “Google it” diminishes. This ambient integration could dramatically boost LLM query volume by the late 2020s, accelerating the crossover. Economic and Content Ecosystem: One challenge is the sustainability of the web content ecosystem. Traditional search drives traffic to websites; AI answers often quote information without a click-through, which has already led to 60% of Google searches ending with no click. If publishers restrict content access or if regulations intervene (to ensure AI tools aren’t anti-competitive), it could impact AI search growth. Conversely, if new monetization models (like AI-native ads or affiliate links in answers) are implemented, AI search could scale faster. By 2030, the advertising and revenue model for search will likely be reinvented to accommodate AI – e.g. sponsored chatbot responses – which could further tilt business incentives toward LLM-based search. Competition and Default Habits: Google’s response will affect the timeline. Google may push its own AI mode (Bard/SGE) to all users by default. If Google successfully retains users within its ecosystem by offering the best of both worlds (trusted AI answers with the option of traditional results), the “overtaking” might be less visible as a Google vs. ChatGPT battle – instead, Google’s search itself becomes LLM-powered. In that case, the tipping point could arrive as Google’s search product transforms into an LLM-first experience by 2030, effectively meaning LLM search has overtaken the old link-based search within the dominant platform. On the other hand, if an independent AI provider (OpenAI or others) captures a large user base directly, that would mark a more distinct overtaking of Google. Current signals (e.g. OpenAI’s plan for a search engine, and ChatGPT becoming a household name) suggest a real possibility of an external AI platform rivalling Google’s scale by 2030. 2030 Is When AI Search Takes the Crown All indicators point to a transformative shift in how people search for information over the next 5-7 years. By 2030, LLM-powered search is projected to eclipse traditional search engines in global usage – a historic changing of the guard in consumer technology. We expect the crossover around 2028-2030, when more daily queries worldwide go through AI assistants than through keyword searches. This will be driven by LLMs’ continued exponential adoption, improvements in AI capabilities, and user preference for convenient, conversational answers. Notably, “overtaking” does not mean search engines vanish overnight – rather, they will evolve or integrate these AI capabilities. In fact, by 2030 the distinction between an “LLM-based assistant” and a “search engine” may blur, as most search platforms will have become AI-centric. In practical terms, the milestone to watch is when LLM-based systems account for >50% of search queries and traffic. Current data and forecasts suggest this is likely by the end of this decade (around 2030), with some metrics reaching parity even sooner (e.g. half of informational searches via AI by 2028). The transition is already underway: users are dividing their searches, businesses are adapting SEO for AI, and search giants are reinventing themselves as AI companies. The adoption curves are on a collision course, and if present trends hold, 2030 is set to be the year LLM-powered search becomes the new dominant paradigm. Sources: The projections and data above are drawn from a range of authoritative sources, including analyst reports, consumer surveys, and public disclosures by the companies involved. Key references include SparkToro’s 2024 search volume research, Gartner’s AI adoption forecasts, Kevin Indig’s industry analysis, and usage statistics from OpenAI and others. These provide a robust, evidence-based foundation for predicting when and how LLM-based search will overtake traditional search in the coming years. Prepare Your Business for the AI Search Era The shift from traditional search to AI-first platforms is accelerating — and the tipping point may arrive sooner than most forecasts suggest. Organizations that act now can adapt their SEO strategies, optimize content for AI-driven discovery, and integrate LLM-powered tools into daily operations. TTMS supports companies worldwide in leveraging AI technologies, automating critical workflows, and ensuring their digital presence remains competitive in the new search landscape. Let’s explore how your business can lead — not follow — in the AI search era. Talk to our experts! Will ChatGPT completely replace Google Search by 2030? While forecasts suggest ChatGPT and other AI-powered assistants could surpass Google in global search share by 2030, complete replacement is unlikely. Instead, search is expected to evolve into a hybrid model where AI tools handle most complex and conversational queries, while traditional engines remain relevant for quick facts, local information, and transactional searches. How will AI search change SEO strategies? AI search shifts the focus from ranking for keywords to being cited as a trusted source within AI-generated answers. This means optimizing content for clarity, authority, and relevance to AI models, while also monitoring “share of voice” in AI responses. Businesses will need to adapt by creating content formats that AI tools can easily summarize and reference. Is AI-powered search more accurate than traditional search engines? Accuracy depends on the query type. For in-depth, multi-step, or creative tasks, AI assistants like ChatGPT often provide richer, more contextual responses. However, for real-time, fact-based queries, traditional engines with live indexing still hold an advantage — though this gap is narrowing as AI integrates real-time data sources. What industries will benefit most from the rise of AI search? Sectors requiring personalized advice, problem-solving, or detailed explanations — such as education, healthcare, travel, software development, and legal services — stand to gain the most. These industries can leverage AI search to deliver tailored recommendations and solutions directly to users without multiple clicks. How can businesses prepare for the AI search tipping point? Companies should start by auditing their content for AI-readiness, ensuring it’s authoritative, well-structured, and easy for AI to parse. They should also monitor how often their brand appears in AI responses, experiment with conversational content formats, and integrate AI tools into customer-facing workflows to stay competitive in the evolving search landscape.

AML Risks in Real Estate: How Automation Helps Reduce Exposure

The real estate sector has long been a target for money laundering with high-value transactions and less oversight than banks. As regulations tighten, property professionals face expanding Anti-Money Laundering (AML) obligations. This article examines key AML requirements under European Union (EU) law, common money laundering risks in real estate, the challenges of manual compliance, and how automation can help firms reduce their exposure. AML Compliance Obligations in Real Estate Under EU law, real estate businesses are considered “obliged entities” subject to AML rules alongside banks. This means real estate professionals involved in property deals must conduct thorough customer due diligence on buyers and sellers. They must verify each client’s identity and determine the ultimate beneficial owner (UBO) for corporate buyers, ensuring the true buyer is known. Clients are also screened against sanctions and politically exposed person (PEP) watchlists to flag any high-risk parties for closer review. In addition, EU regulations mandate ongoing monitoring of the business relationship and reporting of suspicious transactions to the Financial Intelligence Unit (FIU). Real estate firms must report any suspected money laundering to authorities. Regulations also impose strict record-keeping, so firms must document all checks and retain records for years. To curb illicit cash, the EU caps large cash payments (often around €10,000 for property deals). Deals involving high-risk countries or PEPs require enhanced due diligence. Failing to meet these obligations can lead to severe penalties and reputational damage. Common AML Risks in Real Estate Criminals exploit real estate through various money laundering techniques, including: Shell and front companies: Criminals often purchase property via shell companies or complex corporate structures to hide the true owner’s identity. Layers of entities and nominee buyers make it difficult to trace the real source of funds. Large cash transactions: High-value property deals done in cash let criminals inject illicit funds without leaving an obvious audit trail. By avoiding banks, they bypass financial reporting systems, making detection harder. Opaque ownership structures: Trusts and third-party nominees are used to conceal the real owner. Such arrangements hide the ultimate beneficial owner and complicate due diligence. Price manipulation and collusion: Property values can be deliberately over- or under-valued to facilitate laundering. With collusion from complicit insiders, criminals can inflate or deflate prices to disguise illicit money flows. Challenges of Manual AML Compliance For many real estate firms, meeting AML requirements manually is a resource-intensive and error-prone endeavor. Key challenges include: Labor-intensive processes: Performing due diligence for each client means collecting and verifying numerous documents (IDs, proof of funds, etc.) and cross-checking databases. Doing all of this by hand for every transaction is time-consuming and can slow down deals. Risk of human error: Manual processes are prone to oversight and inconsistency. A busy agent might miss a red flag or overlook that a client is on a sanctions list. Such slips can result in compliance violations, regulatory fines, or reputational harm. Keeping up with regulations: AML rules and watchlists change frequently. Without automation, compliance teams must constantly update their knowledge and procedures. Smaller agencies may especially struggle to adjust processes for each new law or directive. Operational delays and costs: Lengthy KYC checks frustrate clients and may even drive them away. Staff hours spent on repetitive checks also raise costs. In short, heavy reliance on manual compliance can hinder business growth while still leaving potential gaps in risk coverage. How Automation Helps Reduce AML Risk Automation can significantly strengthen AML compliance in the real estate sector by addressing many of the above challenges. By leveraging specialized software and data tools, companies can streamline their anti-money laundering efforts. Key benefits of automating AML include: Automated risk assessment: Smart AML platforms automatically risk-rate clients and transactions using predefined criteria. Factors such as a client’s location, profile, and transaction size are analyzed to assign a risk score, flagging high-risk cases for closer review. This ensures attention goes to the areas of highest risk, and the system can continuously monitor for any new suspicious indicators. Digital identity verification: Automation accelerates identity checks by using digital KYC solutions. These tools can scan and authenticate IDs or passports, use biometric verification to confirm the individual, and cross-check clients instantly against sanctions and PEP databases. This not only speeds up onboarding but also reduces the chance of overlooking a high-risk or blacklisted individual. UBO discovery and verification: Modern AML solutions quickly pinpoint ultimate beneficial owners by integrating with global corporate registries. Instead of staff manually untangling complex company ownership, the software reveals who ultimately controls a corporate buyer (for example, any person with over 25% ownership). This makes it much harder for criminals to hide behind layers of companies. Streamlined documentation and reporting: An automated system maintains a complete digital audit trail of all compliance steps. Every ID check, risk score, and ownership verification is logged and stored, making it easy to demonstrate compliance during audits. If a suspicious transaction needs to be reported, the system can help compile the necessary data for regulators, making filings faster and more accurate. Automated AML Solution for Real Estate AMLTrack is software designed for obliged entities such as real estate agencies and property firms. The system automates key AML tasks – from verifying client identities and screening them against sanctions lists and PEP databases, to retrieving data from official registers (KRS – National Court Register, CEIDG – Central Register and Information on Economic Activity, CRBR – Central Register of Beneficial Owners), performing risk assessments, and reporting suspicious transactions. All actions are documented and stored in a secure archive, ready for regulatory inspection. This allows real estate professionals to meet legal requirements more quickly and reliably, reducing the risk of human error and costly compliance breaches. Conclusion Money laundering threats in real estate continue to evolve, but so do the tools to counter them. By embracing automation, property professionals can stay ahead of criminals and meet their AML obligations with greater ease. Ultimately, automated compliance helps firms reduce risk exposure, protect their reputation, avoid hefty fines, and contribute to a more transparent and secure real estate market. Why is real estate considered high-risk for money laundering? The real estate sector is attractive to money launderers because it allows large sums of money to be moved discreetly and converted into stable, long-term assets. Properties—especially in luxury or commercial segments—can be bought with illicit funds and later sold to generate “clean” money. Criminals often use complex structures such as shell companies, nominee buyers, or third parties to mask their identity. Additionally, property valuations can be manipulated to conceal illegal profits. Compared to the banking sector, real estate historically had weaker oversight, making it a soft target for illicit financial activity. Do small real estate agencies also need to comply with AML regulations? Yes. Under EU law, all real estate professionals involved in property transactions are considered “obliged entities.” This includes large commercial developers, small agencies, and even independent brokers who help clients buy or sell real estate. The law does not differentiate based on company size. All entities must conduct customer due diligence, report suspicious activity, and maintain proper compliance documentation. Smaller firms, while often with limited resources, are still subject to the same regulatory scrutiny and risk of penalties for non-compliance. That’s why many turn to automation to streamline their obligations without adding headcount. What is UBO verification and why does it matter in real estate? UBO stands for Ultimate Beneficial Owner — the person who ultimately owns or controls a legal entity. In real estate, it’s crucial to identify the UBO when a property is purchased through a company, trust, or intermediary. Criminals often use multi-layered company structures across jurisdictions to hide the real buyer and the source of funds. By verifying the UBO, real estate firms help prevent anonymous property purchases used to launder money. EU regulations require firms to conduct UBO checks and to apply enhanced due diligence if the ownership structure appears unusually complex or obscured. What are the penalties for non-compliance with AML in real estate? Penalties for AML non-compliance in the real estate sector can be severe. Financial penalties vary by country, but they often reach into the hundreds of thousands or even millions of euros. In some cases, firms may face operational sanctions such as suspension of licenses or exclusion from public contracts. Individuals—such as managing directors or compliance officers—can also be held personally liable if AML failures are found to be due to negligence. Beyond regulatory action, firms risk reputational damage, loss of clients, and negative media coverage. A single lapse in due diligence can have long-term consequences for the business. Can AML automation help with cross-border real estate deals? Yes, AML automation is particularly useful for cross-border transactions, which carry higher risks due to differing legal standards, unfamiliar jurisdictions, and language barriers. Automated platforms can instantly access international databases, perform multilingual identity verification, and screen parties against global sanctions and PEP lists. They can also streamline the collection and validation of documents from foreign clients. This ensures consistency and accuracy while reducing delays, which are common in manual processes. For international property firms and clients, automation provides both operational efficiency and a much stronger compliance posture.

Wiktor Janicki

We hereby declare that Transition Technologies MS provides IT services on time, with high quality and in accordance with the signed agreement. We recommend TTMS as a trustworthy and reliable provider of Salesforce IT services.

Julien Guillot Schneider Electric

TTMS has really helped us thorough the years in the field of configuration and management of protection relays with the use of various technologies. I do confirm, that the services provided by TTMS are implemented in a timely manner, in accordance with the agreement and duly.

Let’s talk about how TTMS can help.

Monika Radomska

Sales Manager

Artificial Intelligence in Pharma: Software Implementation and Validation According to New EU Regulations

1. AI Regulations in the EU 2025: The Artificial Intelligence Act (AI Act)

2. Practical Aspects of AI Implementations in Pharmaceutical Manufacturing: EudraLex Annex 22 Guidelines

3. AI in Drug Manufacturing: Applications and Benefits

4. Ensuring Pharmaceutical Product Quality with Artificial Intelligence

5. AI Implementation and Validation According to New Guidelines: Practical Aspects

6. Summary Ai Software in Pharma Industry

6. How TTMS can help you leverage AI in pharmaceuticals

Recommended articles

TTMS blog – the world through the eyes of IT experts

Artificial Intelligence in Pharma: Software Implementation and Validation According to New EU Regulations

AI in Pharma: Reshaping the Future of Research and Development

6AMLD and the New EU AML Regulation: What It Means for Business

AML in the Art Market: Automation for Safe and Transparent Transactions

LLM-Powered Search vs Traditional Search: 2025-2030 Forecast

AML Risks in Real Estate: How Automation Helps Reduce Exposure

Ready to take your business to the next level?

Let’s talk about how TTMS can help.

Monika Radomska